Model Online / On-Site
Duration
Modules 5 modules

Course Objective & Scope

Public LLM leaderboards rarely tell you how a model will perform on your organisation's Turkish documents, internal API schemas or available GPU budget. This course provides a systematic method for comparing model families such as Llama, Qwen, Mistral, Gemma and DeepSeek across architecture (dense versus MoE), parameter scale, long-context behaviour and commercial licensing terms.

Participants shortlist candidate models for their own use cases, run evaluations across accuracy, Turkish fluency, structured output/tool calling and hardware cost, and produce a defensible model selection report.

Who Should Attend?

  • AI engineers and data scientists making model decisions for internal projects
  • Software and solution architects
  • Technical product managers evaluating a move to open-weight models
  • Prerequisite: basic LLM concepts and calling APIs from Python

Key Takeaways

  • Classify open-weight model families by architecture, scale and specialisation
  • Assess the commercial and regulatory implications of Apache 2.0, MIT and custom licences
  • Analyse the VRAM and speed trade-offs between dense and Mixture-of-Experts (MoE) models
  • Run Turkish-language, RAG and tool-calling accuracy tests on real company tasks
  • Build a model decision matrix combining cost, quality and latency criteria

Curriculum

The Open-Weight Model Landscape
  • Strengths and trade-offs of Llama, Qwen, Mistral, Gemma and DeepSeek families
  • Base, instruct, reasoning (thinking), coder and multimodal variants
Architecture and Hardware Impact
  • Total versus active parameters and VRAM needs in dense and MoE models
  • GQA/MLA attention designs and practical limits of long-context performance
Licensing and Enterprise Compliance
  • Comparing Apache 2.0/MIT with custom Llama, Gemma and Mistral licences
  • User thresholds, derivative model rights and enterprise legal checklist
Task-Specific Benchmarking
  • Pitfalls of public leaderboards and benchmark data contamination
  • Testing Turkish quality, JSON schema adherence and function calling
Model Selection Matrix and Decision Brief
  • Combining small specialist models with larger general-purpose models
  • Selection report unifying quality, latency and GPU cost

Other Courses in This Category

  • Enterprise Model Serving with NVIDIA Triton Inference Server and KServe

    Multi-model GPU serving with NVIDIA Triton Inference Server and KServe: TensorRT-LLM, ONNX and Python backends, dynamic batching, model ensembles, KServe InferenceService, Knative autoscaling and tuning with Model Analyzer.

    Local & On-Premise LLM Platforms Corporate
    Live Course
    5 modules
    View
  • KV Cache Management and Continuous Batching Optimization for LLM Inference

    Solving memory and concurrency bottlenecks in LLM serving: KV cache memory math, PagedAttention, SGLang RadixAttention, prefix caching, FP8 KV cache quantization, continuous batching and chunked prefill tuning.

    Local & On-Premise LLM Platforms Corporate
    Live Course
    5 modules
    View
  • LLM Inference Throughput and Latency Optimization

    Optimising TTFT, ITL and total throughput in LLM services: compute-bound vs memory-bandwidth-bound bottleneck analysis, FlashAttention/FlashInfer, CUDA Graphs, speculative decoding, parallelism tuning and SLO-driven load testing.

    Local & On-Premise LLM Platforms Corporate
    Live Course
    5 modules
    View
  • Running AI and LLM Workloads on Kubernetes

    Operating GPU and LLM workloads on Kubernetes: NVIDIA GPU Operator, GPU sharing with MIG and time-slicing, queue management with Kueue/Volcano, fast model weight caching, GPU autoscaling and DCGM observability.

    Local & On-Premise LLM Platforms Corporate
    Live Course
    5 modules
    View
  • Deploying Small Language Models on Edge and On-Device Hardware

    Running Small Language Models (SLMs) on edge devices, field hardware and NPUs/embedded GPUs: Llama 3.2, Qwen, Gemma and Phi (1B–4B), ONNX Runtime GenAI, ExecuTorch, TensorRT, Jetson and offline field scenarios.

    Local & On-Premise LLM Platforms Corporate
    Live Course
    5 modules
    View
  • LLM Unit Economics: Cost per Request and Token Optimization

    Unit cost analysis and optimisation in LLM systems: self-hosted GPU vs commercial API break-even math, prompt and output token reduction, batch APIs, model cascading and AI FinOps metrics.

    Local & On-Premise LLM Platforms Corporate
    Live Course
    5 modules
    View

Get in touch